📦 romainsimon/uisfx

UI SFX: 936 sounds, zero dependencies

UI SFX is a tiny, zero-dependency sound library that lets you completely change the personality of your app's interface sounds without touching a single line of your logic. Instead of hunting for random audio files, you call semantic cues like 'success' or 'warning', and the system plays the right sound from one of twelve distinct sonic packs. The most surprising part is that you can switch the entire vibe from 'minimal' to 'arcade' in one line of code, while the underlying interaction logic remains exactly the same.

🆔 @hackernewsgithubprojects

📦 thinkany-ai/dscode

DSCode: The Local AI Coder

DSCode is the local coding agent that finally puts you in the driver's seat. It runs right on your machine, keeping all your code and session data private instead of sending it off to some cloud server. The coolest part is its sandboxing. Every command the AI tries to run happens inside a locked-down environment, so even if the model gets a little wild, it cannot touch your actual files or network. You get the speed and low cost of DeepSeek by default, but you can easily swap in other models like OpenAI or Anthropic whenever you need them.

🆔 @hackernewsgithubprojects

📦 r4dl/come

CoMe: The Method That Turns Clouds into Solid Surfaces

CoMe is the mesh extraction tool that finally gives your 3D Gaussian Splatting scenes a solid, usable surface. You know how those new 3D camera techniques create beautiful, realistic images from thousands of tiny, fuzzy splats? The problem is those splats are just a cloud; they are not solid objects you can touch or use in a movie. CoMe solves this by using a smart confidence system to figure out exactly where the solid surface ends and the empty space begins. It is like a digital sculptor that carefully chisels away the noise to reveal a clean, detailed 3D model.

🆔 @hackernewsgithubprojects

📦 internlm/intern-s2-mobius

Intern-S2-Mobius: Faster AI Reasoning

Intern-S2-Mobius is the foundation model that separates memory from thinking to deliver speed. While standard AI systems store facts and process logic layer by layer, this thirty-five billion parameter model uses a shared memory bank that multiple reasoning units query in parallel. This design lets the system refine its thoughts internally rather than writing out long, visible step-by-step notes. The result is a massive four times speedup in inference time because the model thinks more efficiently in the background.

🆔 @hackernewsgithubprojects

📦 vixhal-baraiya/microgpt-c

MicroGPT C: Training a GPT in Pure C

Generate new names by training a character level transformer on a single line of text. MicroGPT C fits an entire neural network into one C file with no external libraries. It uses a tiny model to learn patterns and predict the next letter. The code relies only on your operating system core to handle math and memory. This project proves you can build and run a language model from scratch. You do not need heavy frameworks. You just need a compiler and a text file. The result is a clear look at how AI models actually work under the hood. It is a perfect way to learn the fundamentals.

📰 https://news.ycombinator.com/item?id=49347477

🆔 @hackernewsgithubprojects

📦 subinium/awesome-claude-code

Awesome Claude Code

One repository has become the single most important map for the entire ecosystem of AI coding tools, filtering out the noise to show only projects with over one thousand stars. It is called Awesome Claude Code, and it curates the best plugins, skills, and servers that make these AI assistants actually useful in real workflows. Instead of guessing which tools are worth your time, you get a vetted list that covers everything from security reviews to custom agent orchestration. It solves the problem of information overload by focusing strictly on high-quality, proven tools.

🆔 @hackernewsgithubprojects

📦 solsynth/maidkit

MaidKit: The SSH Manager That Doesn't Touch Your Servers

MaidKit is the SSH toolkit that keeps your servers tidy without installing a single thing on them. It’s a cross-platform app that lets you manage terminals, files, and containers purely over the network, so your infrastructure stays clean and uncluttered. The coolest part? It handles everything from database backups to firewall rules, all while keeping your credentials locked in a secure, encrypted vault. It’s like having a digital assistant that handles the heavy lifting, letting you focus on your code instead of the plumbing. If you want server management that feels calm and capable, this is the tool to keep an eye on.

🆔 @hackernewsgithubprojects

📦 steve245270533/three-stylized

Three Stylized: Instant 3D Meadows

Generate a fully animated, stylized 3D meadow in your browser using Three Stylized. This library handles everything from procedural terrain generation to instanced grass blades, all while keeping performance high. The coolest part is the wind system; it uses custom shaders to make thousands of individual blades sway naturally, creating a living landscape without heavy assets. You can even tweak wind strength, flower density, and lighting colors in real time. It is perfect for anyone wanting to add a lush, interactive nature scene to their web projects without the usual complexity. Just drop it into your scene, call the update loop, and watch the grass dance.

🆔 @hackernewsgithubprojects

📦 earendil-works/pi-review

pi-review

pi-review is the coding agent extension that turns your AI assistant into a disciplined code reviewer. It plugs directly into the Pi coding tool, letting you inspect uncommitted changes, compare against a base branch, or pull in a GitHub pull request for immediate analysis. The standout feature is how it separates the technical feedback given to the AI from the human callouts, ensuring you get actionable follow-ups without the noise. This keeps your workflow focused on what actually matters. By clearly prioritizing findings and offering a definitive verdict, it helps you spot issues before they become problems.

🆔 @hackernewsgithubprojects

📦 deepgrove-ai/mlx-lm-deepgrove

Maple: A 20B AI Model That Runs on Your Laptop

Maple is a twenty billion parameter AI model that fits entirely into about seven gigabytes of memory, which means it runs comfortably on the Apple silicon chip in your laptop. The trick is that it uses a technique called ternary compression, where the model's weights are simplified down to just three states instead of the usual two or four, drastically cutting down the space it needs. This project provides the tools to download and run this specific model on your Mac, letting you chat with a very capable AI locally without needing a powerful server.

🆔 @hackernewsgithubprojects

📦 vistyy/nopus

Nopus Catches When AI Coding Agents Start Babbling

Nopus automatically intercepts coding agent responses that slip into dense, abstract academic language. Instead of letting you read through walls of jargon, this tool measures the readability of your code assistant's text and flags it if it gets too complicated. It then asks the AI to rewrite the answer in plain, direct English. You can adjust how sensitive this check is, ranging from catching only the worst offenders to demanding simple answers every time. The goal is to keep your coding sessions clear and actionable without you having to constantly rephrase your prompts.

🆔 @hackernewsgithubprojects

📦 mengto/sylva

The Moss World That Reacts to Your Mouse

Sylva lets you move through a living, moss-covered landscape that reacts to your cursor as if it were alive. This is a procedural 3D study built with Three.js, where a massive root system grows entirely from code rather than pre-made models. The most surprising part is that the moss parts around your mouse and releases a trail of pollen as you move, making the static scene feel organic and responsive. It renders up to a quarter million blades of moss in real-time, all generated deterministically so the landscape looks the same on every load.

🆔 @hackernewsgithubprojects

📦 dob323/session-kit

Session Kit: Tame Your AI Coding Chaos

Keep your AI coding agents from crashing when your connection drops. Session Kit gives every Claude Code and Codex session a stable number, a unique color, and a clear status that tells you instantly if it is stuck or just churning. It runs on the host, so your work survives SSH disconnects, and it uses a strict safety check to ensure you never accidentally mess with the wrong session. The coolest part is that you can run multiple subscriptions side by side, giving each agent its own login. This tool is your control center, making sure you always know exactly which brain needs your attention next. Stop guessing, start managing.

🆔 @hackernewsgithubprojects

📦 walkinglabs/learn-harness-engineering

Learn Harness Engineering: Stop Your AI Agents From Failing

AI coding agents are incredibly smart, yet they still fail constantly because they guess at your project's rules instead of knowing them. Learn Harness Engineering is a complete beginner tutorial that fixes this by turning your code folder into a strict instruction set for the AI. It walks you through fourteen lectures and hands-on projects to teach you how to stop your agent from losing track of what it is doing or finishing halfway. This is not about learning to code, but about learning how to manage an AI teammate so it actually gets the job done right.

🆔 @hackernewsgithubprojects

📦 therepanic/openleetcode

Open LeetCode: Test Your Code Locally

LeetCode test cases are usually hidden behind a paywall, but a project called Open LeetCode changes that. It is a local tool written in Haskell that lets you run algorithm challenges right on your own computer. You simply drop in your solution file for any language, and it uses open, visible test suites to judge your code instantly. This removes the guesswork of debugging without knowing exactly which inputs failed. The whole system is transparent because every test and rule lives in the public repository, so you can see exactly how your solution is evaluated.

📰 https://news.ycombinator.com/item?id=49347178

🆔 @hackernewsgithubprojects

📦 nvidia/tensorrt-model-connect

TensorRT Model Connect: The Two-Command AI Pipeline

TensorRT Model Connect is the bridge that turns any PyTorch model into a blazing-fast, production-ready inference engine in just two commands. It completely eliminates the manual headache of optimizing models for your hardware, handling the complex conversion to TensorRT for you, so you can go from a raw model file to a highly optimized, cross-platform runtime without writing a single line of glue code. The tool automatically handles precision tuning, kernel selection, and backend setup, ensuring your models run at their absolute peak performance on your specific GPU. This is the ultimate shortcut for developers who want maximum speed and minimum setup friction.

🆔 @hackernewsgithubprojects

📦 vizart-vj/comfyui-minimax-h3-longmedia

ComfyUI MiniMax H3 LongMedia: Run Long Videos on Small GPUs

ComfyUI MiniMax H3 LongMedia allows you to generate continuous, long-form video and audio on hardware that would normally crash under the weight of a single long pass. The project’s most surprising capability is its geometry-aware memory guard, which prevents your graphics card from freezing by streaming attention layers and compressing data in real-time. This is a set of custom nodes for the ComfyUI platform designed specifically for the MiniMax H3 model. It solves the problem of running high-resolution, multi-clip video generations without needing a massive, expensive workstation.

🆔 @hackernewsgithubprojects

📦 squall01337/mixamo-llm-mocap

Mixamo LLM Mocap: Turn Any Video Into Animation

You can turn any static video into a clean, playable animation for any character, without a single motion capture suit or manual keyframe. Mixamo LLM Mocap uses computer vision to read human movement from the footage, then automatically maps that motion onto your chosen digital rig inside Blender. It even handles two people fighting in the same scene, keeping their movements realistic and grounded. The most surprising part is that an AI agent can run this entire process end-to-end, meaning no human needs to touch the software at all.

🆔 @hackernewsgithubprojects

📦 rizriyz/bohay

Bohay: The Mission Control for Your AI Agents

Bohay is a single three megabyte binary that turns your chaotic terminal tabs into a unified mission control for your AI coding agents. The biggest headache with AI coding is that you end up with ten different windows open, half of which are stuck waiting for permission while you forget about the others. Bohay solves this by tracking every agent in one place and showing you exactly which one is blocked, working, or done. You can jump straight to the agent needing your attention without digging through windows. It even remembers your entire setup, so if your terminal closes, your sessions and layout come back exactly as they were.

🆔 @hackernewsgithubprojects

📦 msllrs/relay

Relay: Dictate Your LLM Prompts

Relay is the macOS menu bar app that finally lets you talk to your AI instead of typing. It solves the annoying problem of building complex prompts by letting you dictate while it automatically grabs anything you copy from your screen. You do not have to hunt for files or switch tabs. Just speak, and your notes, code snippets, and links assemble themselves into a perfect, ready-to-paste prompt. It even understands when you draw a circle on your screen to point out what you mean. The coolest part is that it turns your scattered thoughts into a clean, structured message with zero manual formatting.

📰 https://news.ycombinator.com/item?id=47709858

🆔 @hackernewsgithubprojects

📦 shanliuling/dsh-image-gen

Make Your AI Agent Draw Images

DeepSeek Harness agents can now generate images directly inside the chat window, just like you would with a standard AI assistant. This open-source plugin, dsh-image-gen, bridges that gap by letting your agent automatically call image generation tools and render the results right in the conversation flow. It works with Google Gemini, OpenAI, and other providers, so you don't need to copy prompts to external sites anymore. You simply ask your agent to draw something, and the picture appears instantly. The images are even saved with the chat history, so you can revisit them later.

🆔 @hackernewsgithubprojects